Corrections
When we get something substantively wrong, we fix it and log the fix here; small copyedits live in each article's revision history instead. Found an error? Use the "Report issue" button on the article (no account needed). So far we have received 2 reports and reviewed 1, taking a median of 4.8 days. 1 is waiting, the oldest since September 26, 2026. That is our record rather than a promise. See the verifiability policy for how checking works.
321 articles were corrected in the last 30 days. Every correction note and resolved report is listed below, newest first. A fix applied to several articles with the same note is shown once, listing each article; "See the change" opens that version in the article's revision history.
- editorial
Independently reverified Alibaba Group against primary filings, regulator records, academic papers, and a strong secondary source; correct fiscal 2026 figures and the current four-segment structure, remove unsupported or volatile claims, and add explicit citation URLs.
- editorial
Replace the mis-scoped AI product catalog with a sourced software-development life-cycle overview; use the canonical SWE-bench link.
- editorial
Narrow transfer-learning and checkpoint-resumption claims to direct evidence.
- editorial
Reverify and correct ICLR history, review process, statistics, awards, workshop format, and 2026 review policies.
- editorial
Independently verify and source-harden Support Vector Machine; correct kernel parameter and validity boundaries, chronology, hinge normalization, KKT conditions, optimizer, complexity, multiclass, calibration, application, and citation claims while preserving accurate core material.
- editorial
FAA Part 108 final rule was not published in March 2026; update delivery drone operators
- editorial
Independently verified Stability AI's corporate history, Stable Diffusion provenance, financing and 2024 accounts, model portfolio and licensing, partnerships, leadership, safety record, and litigation through 2026-07-28; corrected creator, licensing, revenue, debt, US Getty-case, and board claims; added canonical internal links
- editorial
Independently verified neural scaling-law definitions, Kaplan and Chinchilla results, replication evidence, cross-domain regimes, precision, emergence, data limits, and theory through 2026-07-28; corrected the Chinchilla PIQA value, citation metadata, formulas, attribution, and unsupported universal, cost, energy, and forecast claims
- editorial
Independently verified Ilya Sutskever's biography, research record, OpenAI governance history, SSI financing and leadership, research views, and recognition through 2026-07-28; corrected unsupported dates, causal framing, arithmetic, funding totals, and legal-record claims
- editorial
Independently verified GPQA construction, subsets, human baselines, evaluation protocols, and validity evidence through 2026-07-28
- editorial
Fact-checked FlashAttention algorithms, versions, support, and scoped benchmarks
- editorial
Add the July 2026 pre-IPO round and update cumulative funding
- editorial
Add the Ant Group-led pre-Series A funding
- editorial
KRW 155 billion Series A (July 2026) replaces the seed-round funding total
- editorial
Independently verified Cohere history, financing, models, products, proposed Aleph Alpha transaction, and litigation through 2026-07-28
- editorial
Fix the VAE formulation, qualify variant and application claims, and add primary-source citations
- editorial
Independently verify and source-harden Pre-training; correct scope, history, objectives, transfer, datasets, compute, cost, environmental, bias, and reference-order claims while preserving supported core material.
- editorial
Independently verify and source-harden Gradient Boosting; correct history, loss, histogram-complexity, scikit-learn categorical-version, implementation, benchmark, and interpretation claims while preserving accurate core material.
- editorial
Rebuild ByteDance with dated sourcing; distinguish Moonton agreement from closing, add Seed2.1, reconcile 2026 capex scopes, and clarify 2025 financial tense through 2026-07-28.
- editorial
Independently reverified KV caching through the July 28, 2026 cutoff; adds a substantive lead, corrects the Llama 2 13B payload to 800 KiB per token and 3.125 GiB at 4,096 tokens under explicit FP16 assumptions, distinguishes saved projection work from context-dependent dense attention, corrects final DistServe results, bounds product prompt caching and disaggregation claims, and replaces named footnotes and mutable ecosystem catalogs with thirty primary, peer-reviewed, or version-pinned sources.
- editorial
Distinguish Atlas generations and research configurations; replace unsupported specifications and deployment counts; add the 2026 production status; normalize citations and canonical links.
- editorial
Independently reverified first-generation Sora through the July 28, 2026 cutoff; tightens the exact early-access cohort, launch-tool description, Physics-IQ image-to-video switch-frame protocol, human production credits, and reverse-video-search terminology while preserving the independently reviewed scope.
- editorial
Independently reverified Claude 3.5 Sonnet from 20 primary, peer-reviewed, and government sources; distinguishes its June and October snapshots, corrects the 8,192-token output chronology and pre-existing tool use, removes unsupported architecture, training, market, and Claude 3.5 Opus claims, and adds protocol-bounded benchmark, safety, and lifecycle evidence.
- editorial
Independently reverified ReLU from 26 primary, peer-reviewed, and official sources; corrects chronology, bounds the AlexNet speed result and vanishing-gradient claims, clarifies dead-unit and sparsity behavior, and removes two links to nonexistent targets.
- editorial
Independently reverified Physical AI terminology, research boundaries, system architecture, learning evidence, deployment scope, safety standards, and regulation; removed unsupported forecasts and promotional claims.